diff --git a/headroom/backends/anyllm.py b/headroom/backends/anyllm.py
index 468d20450..1af994e47 100644
--- a/headroom/backends/anyllm.py
+++ b/headroom/backends/anyllm.py
@@ -25,6 +25,44 @@ except ImportError:
     AnyLLM = None  # type: ignore
 
 
+def _convert_anthropic_tool(tool: dict[str, Any]) -> dict[str, Any]:
+    """Convert an Anthropic tool definition to the OpenAI function shape.
+
+    any-llm speaks OpenAI, so an Anthropic ``{name, description, input_schema}``
+    tool must become ``{type: function, function: {name, description,
+    parameters}}`` before it is forwarded, or the provider ignores/rejects the
+    tools array and the model never calls a tool. Mirrors the LiteLLM backend's
+    converter so both OpenAI-compatible backends send the same shape.
+    """
+    func: dict[str, Any] = {"name": tool.get("name", "")}
+    if "description" in tool:
+        func["description"] = tool["description"]
+    if "input_schema" in tool:
+        func["parameters"] = tool["input_schema"]
+    return {"type": "function", "function": func}
+
+
+def _convert_tool_choice(choice: Any) -> Any:
+    """Convert an Anthropic ``tool_choice`` to the OpenAI shape (mirrors LiteLLM).
+
+    Anthropic: ``{"type": "auto"}``, ``{"type": "any"}``, ``{"type": "tool",
+    "name": ...}``. OpenAI: ``"auto"``, ``"required"``, ``{"type": "function",
+    "function": {"name": ...}}``. Passing the raw Anthropic dict through makes
+    the provider reject or ignore it.
+    """
+    if isinstance(choice, str):
+        return choice
+    if isinstance(choice, dict):
+        choice_type = choice.get("type", "auto")
+        if choice_type == "auto":
+            return "auto"
+        if choice_type == "any":
+            return "required"
+        if choice_type == "tool":
+            return {"type": "function", "function": {"name": choice.get("name", "")}}
+    return "auto"
+
+
 class AnyLLMBackend(Backend):
     """Backend using any-llm for multi-provider support."""
 
@@ -251,9 +289,9 @@ class AnyLLMBackend(Backend):
             if "stop_sequences" in body:
                 kwargs["stop"] = body["stop_sequences"]
             if "tools" in body:
-                kwargs["tools"] = body["tools"]
+                kwargs["tools"] = [_convert_anthropic_tool(t) for t in body["tools"]]
             if "tool_choice" in body:
-                kwargs["tool_choice"] = body["tool_choice"]
+                kwargs["tool_choice"] = _convert_tool_choice(body["tool_choice"])
 
             logger.debug(f"any-llm request: provider={self.provider}, model={original_model}")
 
@@ -301,9 +339,9 @@ class AnyLLMBackend(Backend):
             if "stop_sequences" in body:
                 kwargs["stop"] = body["stop_sequences"]
             if "tools" in body:
-                kwargs["tools"] = body["tools"]
+                kwargs["tools"] = [_convert_anthropic_tool(t) for t in body["tools"]]
             if "tool_choice" in body:
-                kwargs["tool_choice"] = body["tool_choice"]
+                kwargs["tool_choice"] = _convert_tool_choice(body["tool_choice"])
 
             msg_id = f"msg_{uuid.uuid4().hex[:24]}"
 
diff --git a/tests/test_backend_anyllm.py b/tests/test_backend_anyllm.py
index 3d56fcb0a..3e1e1a8e1 100644
--- a/tests/test_backend_anyllm.py
+++ b/tests/test_backend_anyllm.py
@@ -293,6 +293,81 @@ async def test_send_message_builds_anthropic_response(monkeypatch: pytest.Monkey
     assert instance.calls[0]["stop"] == ["END"]
 
 
+@pytest.mark.asyncio
+async def test_send_message_converts_anthropic_tools_and_tool_choice(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    """Anthropic tools/tool_choice must reach any-llm in the OpenAI shape.
+
+    any-llm speaks OpenAI; forwarding the raw Anthropic ``input_schema`` tool and
+    the ``{"type": ...}`` tool_choice makes the provider ignore or reject them,
+    so the model never calls a tool. Regression for tool use silently not
+    working on the any-llm backend.
+    """
+    backend, instance = make_backend(monkeypatch)
+    instance.response = make_response(make_choice("ok", "stop"))
+
+    await backend.send_message(
+        {
+            "model": "claude",
+            "messages": [{"role": "user", "content": "hi"}],
+            "tools": [
+                {
+                    "name": "get_weather",
+                    "description": "look up weather",
+                    "input_schema": {
+                        "type": "object",
+                        "properties": {"city": {"type": "string"}},
+                    },
+                }
+            ],
+            "tool_choice": {"type": "any"},
+        },
+        {},
+    )
+
+    sent = instance.calls[0]
+    assert sent["tools"] == [
+        {
+            "type": "function",
+            "function": {
+                "name": "get_weather",
+                "description": "look up weather",
+                "parameters": {"type": "object", "properties": {"city": {"type": "string"}}},
+            },
+        }
+    ]
+    assert sent["tool_choice"] == "required"
+
+
+@pytest.mark.asyncio
+async def test_stream_message_converts_anthropic_tools_and_tool_choice(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    """The streaming request path converts tools/tool_choice the same way."""
+    backend, instance = make_backend(monkeypatch)
+    instance.response = FakeAsyncStream([])
+
+    _events = [
+        event
+        async for event in backend.stream_message(
+            {
+                "model": "claude",
+                "messages": [],
+                "tools": [{"name": "t", "input_schema": {"type": "object"}}],
+                "tool_choice": {"type": "tool", "name": "t"},
+            },
+            {},
+        )
+    ]
+
+    sent = instance.calls[0]
+    assert sent["tools"] == [
+        {"type": "function", "function": {"name": "t", "parameters": {"type": "object"}}}
+    ]
+    assert sent["tool_choice"] == {"type": "function", "function": {"name": "t"}}
+
+
 @pytest.mark.asyncio
 async def test_send_message_returns_error_response(monkeypatch: pytest.MonkeyPatch) -> None:
     backend, instance = make_backend(monkeypatch)